A selective recursive kernel learning-based (SRKL) adaptive predictive controller is proposed for nonlinear time-varying processes. First, a SRKL identification model is presented with an efficient sparsification strategy which makes a trade-off between the tracking precision and the controller’s complexity. The SRKL model can be updated efficiently by introducing and/or deleting a sample via recursive learning algorithms. Consequently, the model can adjust its structure adaptively to capture the process dynamics and time-varying characteristics. On the basis of the SRKL model, a predictive controller with an adaptive modification item is designed. The novel controller can achieve better performance since the SRKL model can trace the process characteristics online. The obtained results on a laboratory-scale liquid-level process and a continuous bioreactor with time-varying parameters show that the proposed controller is superior to the traditional proportional-integral-derivative (PID) controller and related controller with an offline KL model without online updating.
Available at: http://works.bepress.com/inter_liu/8/